Game Forms versus Social Choice Rules as Models of Rights

advertisement
Game Forms versus Social Choice Rules as Models of Rights
Peter J. Hammond
Department of Economics, Stanford University, CA 94305-6072, U.S.A.
A significant revision of the paper presented to the International Economic Association
Round Table on the Theory of Social Choice, held at the Schloss Hernstein, near Vienna,
26–28 May 1994.
ABSTRACT
The paper begins by defining multi-valued game forms which generalize social choice rules.
It also defines the effectiveness relations that represent the rights induced by such game
forms. When one-way instead of two-way rights are allowed, it also demonstrates yet
further difficulties in finding a social choice rule to respect all rights. Finally, to deal with
the problem that game forms typically contain arbitrary features of no consequence to a
society or to its individual members, it is also suggested that one should define both social
choice and individual values over sets consisting of “consequentially equivalent” classes of
strategic game forms.
Modelling Rights
1.
Introduction
When Amartya Sen (1970a) first introduced the idea of rights into formal social
choice theory, he did so through the fairly standard apparatus of a social choice rule (or
SCR) — see also Sen (1970b). By contrast, especially since the provocative work of Peter
Gärdenfors (1981), Robert Sugden (1981, 1985, 1986), and others on this issue, more
recent writers have often preferred to consider game forms. Some of the relationships,
as well as the advantages and disadvantages of each approach, have also been discussed
recently by Riley (1989, 1990), Gaertner, Pattanaik, and Suzumura (1992), Pattanaik and
Suzumura (1992), Sen (1992), and Hammond (1994a).
Rather than go over all these arguments once again, I will suggest a generalized model
of rights that embeds both the SCR and game form approaches as special cases. Indeed,
there is a sense in which the multi-valued game form model I propose shows that “direct”
game forms, in which individuals’ strategies are direct reports of their preferences, are
really implicit in Sen’s original SCR model of rights. After some notation and definitions
have been set out in the preliminary Section 2, this is the topic of Section 3 of the paper.
Next, Sections 4 and 5 discuss some of the different ways in which a multi-valued
game form can be said to respect rights, or to give individuals the power to exercise
their rights. Indeed, this is probably the major difference between the SCR and game
form approaches. Specifically, for arbitrary multi-valued game forms, Section 5 suggests
defining rights in terms of “effectiveness relations.” These generalize the effectiveness
functions which Rosenthal (1972) first introduced into cooperative game theory. Note that
Gärdenfors (1981) also modelled rights in terms of what could be regarded as effectiveness
functions.
In Hammond (1982, 1994a), it was shown how the restriction to “privately oriented”
preferences allows “two-way” rights to be respected. Moreover, the conflict between respect for rights and Pareto efficiency disappears entirely. Indeed, as Coughlin (1986) has
pointed out, Pareto efficiency even requires respect for rights in this special case. But
this is only because, by assumption, no externalities arise in the exercise of individual or
group rights. Section 6 notes how this particular restriction leads to severe difficulties
when individuals and groups are both allowed to have one-way rights.
1
Accordingly, except in special cases, society has no way of avoiding the choice of which
rights to respect and which to violate. Now, one can think of choosing a configuration of
rights as being equivalent to choosing a game form that induces those rights. In the past, an
objection to this approach has been that the typical (strategic) game form includes several
arbitrary features, such as the names of the strategies, etc. To avoid this arbitrariness,
Section 7 considers equivalence classes of “consequentialist” reduced strategic game forms,
along the lines of invariance requirements in classical normal form game theory. In the
end it appears reasonable to consider both social choice and individual values as being
defined over sets of such equivalence classes.
Finally, Section 8 contains a brief concluding assessment.
2.
Preliminary Notation and Definitions
As usual in social choice theory, I shall suppose that there is a fixed domain or
underlying set X of conceivable social states, and a fixed finite set of individuals N with
variable preference orderings Ri (i ∈ N ), which are members of the set R(X) of all
logically possible complete and transitive binary relations defined on X. Let Pi and Ii
(i ∈ N ) denote the corresponding strict preference and indifference relations, respectively;
these must also be transitive. A preference profile RN = Ri i∈N is a listing of preference
orderings, one for each individual in society. Such a profile belongs to the Cartesian
product space RN (X) := i∈N Ri (X), where each Ri (X) is a copy of R(X).
In social choice theory, it is usual to let the feasible set A range over the whole space
F(X) of all possible non-empty finite feasible subsets of X. Here, however, I shall follow the
mechanism design literature, and specifically Dasgupta, Hammond, and Maskin (1979), in
treating the feasible set A ∈ F(X) as fixed. A social choice rule (SCR) therefore specifies,
for the given feasible set A and for each preference profile RN ∈ RN (X), a non-empty
subset CA (RN ) ⊂ A.
For each preference profile RN and each non-empty G ⊂ N , let PG (RN ) and PG∗ (RN )
denote the corresponding strict and strongly strict group preference relations defined for
all pairs a, b ∈ X by
a PG (RN ) b ⇐⇒ ∀i ∈ G : a Pi b
and a PG∗ (RN ) b ⇐⇒ {[ ∀i ∈ G : a Ri b] & [∃h ∈ G : a Ph b ]}.
2
Because each individual’s preference relation is transitive, so are the relations PG (RN )
and PG∗ (RN ). In particular, the weak and the strict Pareto dominance relations PN (RN )
and PN∗ (RN ) are both transitive.
Finally, two further pieces of notation. First, given any Cartesian product set of
the form V N = i∈N Vi and any subset G ⊂ N , write V G for the restricted Cartesian
product set i∈G Vi , with typical member v G = vi i∈G . Second, write V−i for the
product set V N \{i} = h∈N \{i} Vh with Vi omitted, and v−i for the typical member
v N \{i} = vh h∈N \{i} of V−i .
3.
Social Choice Rules and Multi-valued Game Forms
Sen’s (1970a, b) original representation of minimal liberty took a highly specific form.
In a society with a finite set of individuals N , it requires at least two individuals j, k ∈ N
to be given the right to be decisive over at least one personal issue each in the form of
a single pair of alternative social states — say, aj versus bj , and ak versus bk . Here, the
subset { aj , bj , ak , bk } ⊂ A has to have at least three distinct members. Thus, there has
to be an SCR which respects at least the minimal rights of j and k in the sense that:
aj Pj bj =⇒ bj ∈ CA (RN );
ak Pk bk =⇒ bk ∈ CA (RN ).
Sen showed how such minimal liberty was incompatible with even weak Pareto efficiency, in general. Then Batra and Pattanaik (1972) demonstrated a similar incompatibility for “minimal federalism,” when different groups were given the right to determine
issues by the unanimous expression of a strict preference. Not long afterwards, Gibbard
(1974) showed how it might be impossible to respect individual rights at all, even if one
did not insist on Pareto efficiency.
Though at the time they seem not to have been considered as such, the SCRs used
in this early work could be regarded as multi-valued extensions of direct mechanisms, in
the terminology we used in Dasgupta, Hammond and Maskin (1979) that seems to have
become widely accepted since. That is, the individuals i ∈ N could be regarded as players
of a multi-valued game form whose strategies are direct reports of their own preferences.
The SCR considered by Sen then describes what would happen, for each fixed feasible set
A ∈ F(X), if preferences were reported sincerely.
Gibbard (1974) also considered what would happen in a game where individuals could
make strategic decisions to waive their rights in some instances. Actually, Gibbard’s work
3
served to suggest more than just how strategic aspects could play a role when individuals
were accorded certain rights. It also suggested the need to consider indirect mechanisms,
meaning extended game forms in which the strategy sets could be more general than
reports of preferences. This, along with Fine’s (1975) early exploration of the relationship between Sen’s paradox and prisoners’ dilemma, may have done much to prompt the
much more explicit game-theoretic approach to be found in later works such as those by
Gärdenfors (1981), Sugden (1981, 1985, 1986), and Seidl (1986).
single-valued
multi-valued
direct
singleton social choice rules
social choice rules
indirect
game forms
multi-valued game forms
Table 1. The four kinds of social choice rule or game form.
As a way of generalizing both the SCR and the game form models of rights, I shall
follow a suggestion due to Peleg (1984b) for cooperative games. For every fixed feasible
set A ∈ F(X), consider a multi-valued game form N, S N , ΓA (sN ) in which: (i) N is
the set of individual players; (ii) S N := i∈N Si is the set of possible strategy profiles;
(iii) ΓA (sN ) ⊂ A is the non-empty social choice set, defined for every strategy profile
sN = si i∈N ∈ S N . Note that the SCR approach to rights postulates a direct game form,
in which each individual’s strategy set Si consists of the set R(X) of all logically possible
preference orderings over the domain X. The same approach also ignores the incentives
which individuals generally have to misrepresent their true preferences. On the other
hand, the usual game form approach to rights postulates a function with single values
gA (sN ) ∈ A instead of the (generally multi-valued) correspondence ΓA (sN ).
After including the most special case of singleton (direct) social choice rules, the
four different kinds of social choice rule or game form are described in Table 1. Note in
particular that a multi-valued game form is really equivalent to an indirect social choice
rule.
4
4.
Rights-Respecting Social Choice Rules
Given any pair a, b ∈ X, the (non-empty) group G ⊂ N is said to be decisive for a
over b if, whenever a, b ∈ A and the profile RN is such that a PG (RN ) b, then b ∈ CA (RN ).
In other words, group G should be able to veto b if a is available and the members of G
unanimously prefer a to b. The same definition applies to individuals i ∈ N , of course,
taking G = {i}.
It is usual to regard the (one-way) rights of each group G ⊂ N as being represented
by a (possibly empty) collection DG ⊂ X × X of ordered pairs for which G is supposed
to be decisive. Of course, this set DG can be regarded as the graph of a binary preference
relation; this being so, DG can be called a one-way rights relation without undue confusion.
It will be assumed that DG is irreflexive — i.e., that there is no x ∈ X with x DG x.
In case DG is symmetric, it will be called a two-way rights relation. On the whole, the
literature has typically considered two-way rights. The example of Section 7 should serve
to show, however, that this can be unduly restrictive. Accordingly, most results in this
paper require only one-way rights, though the corollary in Section 6 will require two-way
rights relations.
Let G denote the collection of groups G having non-trivial rights relations DG . In
case only individuals have rights, it will be true that G ⊂ { {i} | i ∈ N }. But here I allow
the possibility that groups may have rights, and also that some or all individuals may have
no rights. Usually Di instead of D{i} will be used to indicate individual i’s rights relation.
In what follows, it will be assumed that a particular rights profile DG of irreflexive
rights relations DG G∈G has been specified, for some set G ⊂ 2N of groups (and individuals) with rights. Note that, if G is a proper subset of G, then G will be decisive over
{ a, b } whenever G is.
Finally, say that the SCR CA (RN ) respects the rights profile DG if, whenever a, b ∈ A
with a DG b, then G is decisive for a over b.
5
5.
Rights Induced by Multi-valued Game Forms
Let N, S N , ΓA be a multi-valued game form, with ΓA : S N →
→ A. Also, let G ⊂ N
be a particular group of individuals, and Y, Z ⊂ X a disjoint pair of non-empty sets of
α
Z,
social states. In this case, say that G is α-effective for Y over Z, and write Y DG
iff there exists s̄G ∈ S G such that, for all sN ∈ S N satisfying ΓA (sN ) ∩ Z = ∅, one has
α
Z requires that group G alone always has the
ΓA (s̄G , sN \G ) ⊂ Y . In other words, Y DG
power to veto any outcome in the set Z and restrict the possible social choice outcome to
the set Y instead, no matter what fixed strategies sN \G the other individuals choose. One
α
says then that DG
is group G’s α-rights relation induced by the multi-valued game form.
Suppose that, because of a mistake by the members of group G or for some other
reason, some social state b ∈ Z is still a social outcome that results from the game form,
α
even though Y DG
Z and a Pi b for all i ∈ G and for all a ∈ Y . Yet it is illegitimate for G
to claim any rights violation, because the game form did give G the opportunity to veto
the outcome b.
β
On the other hand, say that G is β-effective for Y over Z, and write Y DG
Z,
iff for all sN ∈ S N satisfying ΓA (sN ) ∩ Z = ∅, there exists s̄G (sN \G ) ∈ S G such that
ΓA (s̄G (sN \G ), sN \G ) ⊂ Y . Unlike with α-effectiveness, β-effectiveness allows the group
G to choose its strategy s̄G (sN \G ) as if responding to the expectation that the other
β
is group
players will choose the appropriate components of sN \G . One says then that DG
β
G’s β-rights relation induced by the multi-valued game form. Evidently Y DG
Z implies
α
Y DG
Z, but the converse is generally false. Of course, this distinction between α- and
β-effectiveness is precisely analogous to the distinction between the α- and β-characteristic
functions of a cooperative game without side payments, as well as to the corresponding
distinction between the α- and β-cores of such a game. See, for example, Aumann and
Peleg (1960) and also Aumann (1967).
β
α
In interpreting Y DG
Z or Y DG
Z as indicating that the game form gives G the right
to Y over Z, it is important that the members of Y and Z consist of completely described
social states or consequences. In particular, everything of relevance to the individuals in
group G must be included in each consequence description. For example, if G’s members
can only avoid the damaging effects of other individuals’ polluting activities by paying
them to desist, then G does not have the right to avoid pollution; in fact, G only has the
right to achieve less pollution at the cost of less money. Similarly, the right to be free
6
of tobacco smoke without having to ask is different from the right to ask others not to
smoke; at least, this is true if non-smokers are troubled by the need to ask, because then
the need to ask becomes part of the relevant consequence.1
Also note that the above definitions are of binary effectiveness relations, rather than
the “effectiveness” or “effectivity functions” considered by Rosenthal (1972), Moulin and
Peleg (1982), Moulin (1983), Peleg (1984a, b), and Deb (1990, 1994).2 One reason for
this is the closer analogy with the binary rights relations for social choice rules considered
by Sen and many later writers. A second reason is greater generality. For, given the
feasible set A ∈ F(X) and the subset Y ⊂ A, observe that G is α-effective for the set Y iff
α
A \ Y , and similarly for β-effectivity. Thus, it is easy to infer effectiveness functions
Y DG
from effectiveness relations, but the converse is not true at all.
β
α
Furthermore, the binary effectiveness relations DG
and DG
are defined without refer-
ence to an equilibrium of any kind. This is crucial. Indeed, were we to consider the rights
to consequences that emerge from an equilibrium, we would be back with direct SCRs
→ S N is an equilibrium correspondence,
and direct rights to consequences. For if E : RN →
then there is an equivalent direct SCR CA : RN →
→ A defined by CA (RN ) := ΓA (E(RN ))
for all RN ∈ RN — cf. Dasgupta et al. (1979).
For the special case when ΓA : S N →
→ A is a single-valued function gA : S N → A
and also Y = {a} and Z = {b}, the above definitions become much simpler. In fact, then
α
{a} DG
{b} iff there exists s̄G ∈ S G such that, for all sN ∈ S N satisfying gA (sN ) = b,
β
{b} iff, for all sN ∈ S N satisfying
one has gA (s̄G , sN \G ) = a. On the other hand, {a} DG
gA (sN ) = b, there exists s̄G (sN \G ) ∈ S G such that gA (s̄G (sN \G ), sN \G ) = a. Indeed,
β
the “induced rights relation” considered in Hammond (1994) is exactly the relation DG
,
applied to single-valued game forms and to individual elements of the feasible set A.
1
I am grateful to Jerry Kelly for a discussion which showed the need to emphasize this point.
Deb also defines “waiver” functions, and gives a much more extensive discussion of how both
effectivity and waiver functions can be used to represent important aspects of rights.
2
7
6.
Rights Oriented Preferences
What rights relations can be respected by a direct social choice rule? A famous
example in the article by Gibbard (1974) was the first to show that not all can be. This
issue was also considered by Farrell (1976). Then Suzumura (1978) defined the profile DN
of individual rights relations to be “coherent” provided that, for an unrestricted domain of
preference profiles RN , there exists an SCR CA (RN ) which respects those rights for every
feasible set A ∈ F(X). However, Gibbard’s example serves to show how very restrictive
is the assumption of coherence.
Thereafter, Blau (1975), Seidl (1975), Farrell (1976), and Breyer (1977) are just a few
of the early works suggesting that restricting the domain of preferences could allow rights
to be respected. In Hammond (1982, 1994a), it was shown that restricting attention to
“privately oriented” preferences allows two-way rights to be respected, and that Pareto
efficiency even requires such rights to be respected. In the case of one-way rights, the
obvious corresponding restriction is the following. Say that individual preferences are
rights oriented provided that, for every i ∈ N , whenever a DG b for some G satisfying
i ∈ G ∈ G, it is true that a Ri b. Thus, i’s rights oriented preferences defer to the rights held
by any group G to which i does not belong by expressing an appropriate weak preference
for G to be allowed to exercise its right. By assumption, therefore, no externalities can
possibly arise in the exercise of individual or group rights — cf. Hillinger and Lapham
(1971).
Of course, in case both a DG b and b DG a for some G satisfying i ∈ G ∈ G, rights
oriented preferences must satisfy a Ii b. This is exactly what is required of privately
oriented preferences when all rights are two-way.
The following is the obvious counterpart of Theorem 1 in Hammond (1994a), which
generalizes a result due to Coughlin (1986):
Theorem 1. Suppose that, for the given rights profile DG , the domain of allowable preference profiles RN is restricted to rights oriented preferences. Then the social choice rule
CA (RN ) satisfies the strict Pareto rule for all A ∈ F(X) only if it respects both individual
and group rights.
Proof: Suppose that the feasible set A ∈ F(X), social states a, b ∈ A, and the group
G ∈ G are such that a DG b. Suppose too that the rights oriented preference profile RN
satisfies a PG (RN ) b. Since RN is rights oriented, it follows that a Ri b for all i ∈ N \ G,
8
and so a PN∗ (RN ) b. If the SCR is strictly Paretian, therefore, it must be true that
b ∈ CA (RN ), proving that G is decisive for a over b. So all rights are respected by any
strictly Paretian SCR.
Indeed, when preferences are rights oriented, then liberalism “does not demand anything that the Pareto principle does not also demand,” as Sen (1971, pp. 1406–7) puts it
in his response to Hillinger and Lapham (1971).
However, in the case of one-way rights, it seems extremely unreasonable to require
preferences to be rights oriented. To show this, I shall use once again a variation of an
example considered by Gibbard (1974, p. 398) and Gärdenfors (1981) which was apparently
suggested by Gilbert and Sullivan’s operetta Trial by Jury. Society N consists of three
individuals — A (Angelina), E (Edwin), and J (the male judge). The domain X consists
of the three social states 0, e, j, where 0 indicates that Angelina marries neither of the two
men, e that she marries Edwin, and j that she marries the judge. It is natural to give
each of the two couples { A, E } and { A, J } the one-way group right to marry, and to
give each individual the one-way right to avoid marrying someone they would rather not.
Thus, the one-way rights relations can be expressed as:
e DAE 0;
j DAJ 0;
0 DA e;
0 DE e;
0 DA j;
0 DJ j.
In this case, it is easily seen that restricting preferences to be rights oriented requires:
0 RA e;
0 RA j;
0 IE j;
0 IJ e.
Hence, in the event that Angelina does not marry him, each man is required to be indifferent about whether she marries the other. This certainly has some ethical appeal.
However, it is deeply troubling that Angelina cannot have any strict preference for marrying either man as opposed to neither of them; this is what follows from not allowing
her to express a strict preference that conflicts with the possible desire of either man not
to marry Angelina. It seems that other ways of satisfying rights while escaping Pareto
inefficiency will have to be found when individuals’ and groups’ one-way rights come into
conflict in this way.
At this point, it is worth emphasizing once again how these and other related negative
results bring out the difficulty there is in satisfying all individual and group rights, even
if appropriate restrictions are placed on individuals’ preferences. It seems inevitable,
9
therefore, that any SCR is bound to respect only some rights, leaving others to be violated.
As Hammond (1994a) and Pattanaik and Suzumura (1992) both discuss, it then becomes
important to have an explicit framework for discussing the social choice of which rights to
respect.3 The SCR approach does allow this, but limits attention to direct multi-valued
game forms, even though the rights induced by an appropriate indirect multi-valued game
form may often be both more extensive and ethically superior to any that an SCR can
respect.
7.
Consequentially Equivalent Strategic Game Forms
As pointed out in Section 1, a typical (strategic) game form includes several arbi-
trary features, such as the names of the strategies, etc. Yet such arbitrariness is easily
circumvented by considering suitable equivalence classes of game forms with the same
“consequentialist” reduced normal form. The concept of “reduced normal form” set out
below will be the weaker version due to Dalkey (1953) rather than the stronger version
used by Thompson (1952) and by Kohlberg and Mertens (1986). An important reason is
that the latter themselves provide (p. 1019, game ∆5 ) an example illustrating that consequentialist reduced normal form invariance, in their stronger sense, is not always possible
when subgame perfect equilibria are being considered.
So, let Γ = N, S N , ΓA (sN ) and Γ̃ = N, S̃ N , Γ̃A (sN ) be two multi-valued game
forms, each with the same set of players N and the same feasible set A of social states.
These two are obviously identical when Si = S̃i for all i ∈ N and ΓA (sN ) = Γ̃A (sN ) for
all sN ∈ S N = S̃ N . But even when these conditions are not true, the two game forms
may effectively be equivalent because one or more players’ strategies have simply been
replicated, or because Γ and Γ̃ differ only in the way in which strategies are labelled. This
motivates the following definitions.
First, the two game forms Γ = N, S N , ΓA (sN ) and Γ̃ = N, S̃ N , Γ̃A (sN ) are said to
be (strategically) equivalent if there is a one-to-one product correspondence ρN : S N ↔ S̃ N
that satisfies ρN (sN ) ≡ i∈N ρi (si ) for some collection ρi : Si ↔ S̃i (i ∈ N ) of individual
one-to-one correspondences, and also has the property that Γ̃A (ρN (sN )) = ΓA (sN ) for all
3
Early on, Hillinger and Lapham (1971) were among those to suggest, at least implicitly,
that such a framework might be necessary. They did not put forward anything like a formal
framework, however.
10
sN ∈ S N . Because ρN is a one-to-one correspondence, this implies that ΓA ((ρN )−1 (s̃N )) =
Γ̃A (s̃N ) for all s̃N ∈ S̃ N .
A second form of equivalence treats all repetitions of “hyper-rows” of the payoff
matrix as irrelevant. So there is an equivalence class of game forms which all share the
same reduced normal form, without any such repetitions. Indeed, suppose the two game
forms Γ = N, S N , ΓA (sN ) and Γ̃ = N, S̃ N , Γ̃A (sN ) are such that Si ⊂ S̃i for all i ∈ N ,
while ΓA (sN ) = Γ̃A (sN ) whenever sN ∈ S N ⊂ S̃ N . Moreover, suppose that for every
player i ∈ N and every strategy s̃i ∈ S̃i \ Si , there exists some alternative si ∈ Si such
that ΓA (si , s−i ) = Γ̃A (s̃i , s−i ) for all s−i ∈ S−i . Then s̃i simply duplicates si in all its
possible consequences, so s̃i may as well be omitted from the strategy set S i . Since this
is true for all s̃i ∈ S̃i \ Si by hypothesis, it follows that Γ̃ is effectively equivalent to Γ.4
With this in mind, consider the obvious way of reducing as far as possible the strategy
sets of the game form Γ = N, S N , ΓA (sN ). Begin by saying that player i’s two strategies
si , s̄i ∈ Si are (consequentially) equivalent if ΓA (si , s−i ) = Γ̃A (s̄i , s−i ) for all s−i ∈ S−i .
This is obviously an equivalence relation, so let Σi be the corresponding collection of
equivalence classes, and write σi (si ) for the unique equivalence class containing the par
ticular strategy si . Let ΣN := i∈N Σi , with typical member σ N . Note that, whenever
σ N (sN ) = σ N (s̄N ) because si and s̄i are in the same equivalence class for each i ∈ N , then
C(A, sN ) = C(A, s̄N ). Hence, there is a well defined outcome function Γ̂A (σ N ) satisfying
Γ̂A (σ N (sN )) := ΓA (sN ) for all sN ∈ S N . The result of this construction is the reduced
normal form Γ̂ = N, ΣN , Γ̂A (σ N ) with the property that, for all i ∈ N and every disjoint
pair σi , σ̄i ∈ Σi , there exists σ−i ∈ Σ−i for which Γ̂A (σi , σ−i ) = Γ̂A (σ̄i , σ−i ).
Finally, the two game forms Γ = N, S N , ΓA (sN ) and Γ̃ = N, S̃ N , Γ̃A (sN ) are said
to be (reduced normal form) equivalent if their corresponding reduced normal forms are
strategically equivalent, for some collection ρi : Σi ↔ Σ̂i (i ∈ N ) of one-to-one correspondences between their respective reduced strategy sets. It is this notion of equivalence that
eliminates irrelevant features such as the re-labelling or duplication of strategies.
4
Deb (1990, 1994), in his discussion of closely related issues, assumes that all such repeated
hyper-rows are simply deleted from the game form.
11
8.
Concluding Assessment
Social systems are game forms, possibly even single-valued, from which social states
or consequences emerge as a result of individuals’ strategic decisions. This paper has
pointed out that multi-valued game forms are rather natural extensions of social choice
rules. It has recalled the difficulties in having a social choice rule respect rights. This
suggests the impossibility of avoiding a choice of what rights to satisfy.
Of course, there are difficulties in modelling rights as game forms. These difficulties
seem greater for multi-valued game forms, which is one good reason why previous writers
have preferred to work with single-valued game forms. Still, one can discuss induced rights
within game forms, and even ask which game form models of a social system induce the
most satisfactory configurations of individual and group rights. This is the approach which
seems to me most worth pursuing in future work on this subject.
Finally, let me return to the main subject and summarize the contrasts between the
two approaches to rights. In game forms, induced rights concern strategies directly, and
social states or consequences only indirectly. Indeed, rights to consequences in game forms
are often contingent, depending on other individuals making suitable strategic decisions.
Whereas in social choice rules, rights are to consequences directly, and these rights are
non-contingent — provided that the social choice rule does manage to respect them, which
can be problematic.
The main difference, therefore, is the indirect and contingent respect for rights that
game forms typically induce, in contrast to the direct and absolute demands of the social
choice approach. Of course, the game form approach is more general, especially when
multi-valued game forms are allowed. The game form approach also permits a greater
variety of ways in which rights can be respected (or violated, of course) — and even an
increase in the set of (contingent) rights which it is feasible to respect.
References
Aumann, R.J. (1967) “A Survey of Cooperative Games without Side Payments”, in
Shubik, M. (ed.) Essays in Mathematical Economics (Princeton: Princeton University
Press), pp. 3–27.
12
Aumann, R.J. and Peleg, B. (1960) “Von Neumann–Morgenstern Solutions to Cooperative
Games without Side Payments”, Bulletin of the American Mathematical Society, vol.
66, pp. 173–179.
Batra, R.N. and Pattanaik, P.K. (1972) “On Some Suggestions for Having Non-Binary
Social Choice Functions”, Theory and Decision, vol. 3, pp. 1–11.
Bernholz, P. (1974) “Is a Paretian Liberal Really Impossible?”, Public Choice, vol. 20,
pp. 99–107.
Blau, J.H. (1975) “Liberal Values and Independence”, Review of Economic Studies, vol.
42, pp. 395–401.
Breyer, F. (1977) “The Liberal Paradox, Decisiveness over Issues, and Domain Restrictions”, Zeitschrift für Nationalökonomie, vol. 37, pp. 45–60.
Coughlin, P.J. (1986) “Rights and the Private Pareto Principle”, Economica, vol. 53, pp.
303–320.
Dalkey, N. (1953) “Equivalence of Information Patterns and Essentially Determinate
Games”, in Kuhn, H. and Tucker, A.W. (eds.) Contributions to the Theory of
Games, Vol. 2: (Annals of Mathematical Studies No. 28) (Princeton: Princeton
University Press), pp. 217–243.
Dasgupta, P.S., Hammond, P.J. and Maskin, E.S. (1979) “The Implementation of Social
Choice Rules: Some General Results on Incentive Compatibility”, Review of Economic
Studies, vol. 46, pp. 185–216.
Deb, R. (1990) “Rights as Alternative Game Forms: Is There a Difference of Consequence?”, preprint, Department of Economics, Southern Methodist University,
Dallas.
Deb, R. (1994) “Waiver, Effectivity, and Rights as Game Forms”, Economica, vol. 61, pp.
167–178.
Farrell, M.J. (1976) “Liberalism in the Theory of Social Choice”, Review of Economic
Studies, vol. 43, pp. 3–10.
Fine, B. (1975) “Individual Liberalism in a Paretian Society”, Journal of Political
Economy, vol. 83, pp. 1277–1281.
13
Gaertner, W., Pattanaik,, P.K. and Suzumura, K. (1992) “Individual Rights Revisited”,
Economica, vol. 59, pp. 161–177.
Gärdenfors, P. (1981) “Rights, Games and Social Choice”, Noûs, vol. 15, pp. 341–356.
Gibbard, A. (1974) “A Pareto-Consistent Libertarian Claim”, Journal of Economic
Theory, vol. 7, pp. 388–410.
Hammond, P.J. (1982) “Liberalism, Independent Rights and the Pareto Principle”, in
Cohen, L.J., L
Y oś, J., Pfeiffer, H. and Podewski , K.-P. (eds.) Logic, Methodology and
the Philosophy of Science, VI (Amsterdam: North-Holland), pp. 217–243.
Hammond, P.J. (1994) “Social Choice of Individual and Group Rights”, in Barnett, W.A.,
Moulin, H., Salles,, M. and Schofield , N. (eds.) Social Choice, Welfare, and Ethics
(Proceedings of the first meeting of the Society for Social Choice and Welfare, Caen,
June 1992) (Cambridge: Cambridge University Press), ch. 3 [to appear].
Hillinger, C. and Lapham, V. (1971) “The Impossibility of a Paretian Liberal: Comment
by Two Who Are Unreconstructed”, Journal of Political Economy, vol. 79, pp.
1403–1405.
Kohlberg, E. and Mertens, J.-F. (1986) “On the Strategic Stability of Equilibria”,
Econometrica, vol. 54, pp. 1003–1037.
Moulin, H. and Peleg, B. (1982) “Cores of Effectivity Functions and Implementation
Theory”, Journal of Mathematical Economics, vol. 10, pp. 115–145.
Pattanaik, P.K. and Suzumura, K. (1992) “Individual Rights and Social Evaluation: A
Conceptual Framework”, University of California at Riverside, Working Paper in
Economics No. 92-29.
Peleg, B. (1984a) Game Theoretic Analysis of Voting in Committees (Cambridge:
Cambridge University Press).
Peleg, B. (1984b) “Core Stability and Duality of Effectivity Functions”, in Hammer, G.
and Pallaschke, D. (eds.) Selected Topics in Operations Research and Mathematical
Economics (Berlin: Springer Verlag), pp. 272–287.
Riley, J. (1989) “Rights to Liberty in Purely Private Matters, Part I”, Economics and
Philosophy, vol. 5, pp. 121–166.
14
Riley, J. (1990) “Rights to Liberty in Purely Private Matters, Part II”, Economics and
Philosophy, vol. 6, pp. 27–64.
Rosenthal, R.W. (1972) “Cooperative Games in Effectiveness Form”, Journal of Economic
Theory, vol. 5, pp. 88–101.
Seidl, C. (1975) “On Liberal Values”, Zeitschrift für Nationalökonomie, vol. 35, pp.
257–292.
Seidl, C. (1986) “The Impossibility of Nondictatorial Tolerance”, Journal of Economics
/ Zeitschrift für Nationalökonomie (Supplementum 5: Welfare Economics of the
Second Best, edited by Bös, D. and Seidl, C.), vol. 46, pp. 211–225.
Sen, A.K. (1970a) “The Impossibility of a Paretian Liberal”, Journal of Political Economy,
vol. 78, pp. 152–157; reprinted in Sen (1982).
Sen, A.K. (1970b) Collective Choice and Social Welfare (San Francisco: Holden Day).
Sen, A.K. (1971) “The Impossibility of a Paretian Liberal: Reply”, Journal of Political
Economy, vol. 79, pp. 1406–1407.
Sen, A.K. (1982) Choice, Welfare and Measurement (Oxford: Basil Blackwell).
Sen, A.K. (1992) “Minimal Liberty”, Economica, vol. 59, pp. 139–159.
Sugden, R. (1981) The Political Economy of Public Choice (Oxford: Martin Robertson).
Sugden, R. (1985) “Liberty, Preference and Choice”, Economics and Philosophy, vol. 1,
pp. 213–229.
Sugden, R. (1986) The Economics of Rights, Co-operation and Welfare (Oxford: Basil
Blackwell).
Suzumura, K. (1978) “On the Consistency of Libertarian Claims”, Review of Economic
Studies, vol. 45, pp. 329–342.
Thompson, F. (1952) “Equivalence of Games in Extensive Form”, U.S. Air Force, Project
RAND, Research memo. RM-759.
15
ACKNOWLEDGEMENTS
Support from an Alexander von Humboldt Foundation research award for a visit to Germany during the academic year 1993–4 is gratefully acknowledged, as is the hospitality
of the University of Kiel where this paper was written. The work it reports was initiated
at the prompting of Amartya Sen. It has also benefited from useful interchanges during
the conference with the discussant Rajat Deb, as well as with Bhaskar Dutta, Jerry Kelly,
Prasanta Pattanaik, and Bezalel Peleg.
Other Endnotes
1. I am grateful to Jerry Kelly for a discussion which showed the need to emphasize this
point.
2. Deb also defines “waiver” functions, and gives a much more extensive discussion of how
both effectivity and waiver functions can be used to represent important aspects of rights.
3. Early on, Hillinger and Lapham (1971) were among those to suggest, at least implicitly,
that such a framework might be necessary. They did not put forward anything like a formal
framework, however.
4. Deb (1990, 1994), in his discussion of closely related issues, assumes that all such
repeated hyper-rows are simply deleted from the game form.
16
Download